Correspondence: Pascal van Hattum, Department of Methodology and Statistics, Faculty of Social Sciences, Utrecht University, PO Box 80140, NL-3508 TC, Utrecht, The Netherlands. Tel: +31 30 2537983; Fax: +31 30 2535797; e-mail: p.vanhattum@uu.nl

¹studied Business Mathematics and Computer Science at the Free University Amsterdam. After his study (2000) he started as a statistical consultant at The SmartAgent Company, where he is currently responsible for all the statistical processes and research & development. In 2004 he started a PhD project in the Faculty of Social and Behavioural Sciences of Utrecht University under the supervision of Professor Herbert Hoijtink. In 2009 he hopes to finalise his thesis, 'Market Segmentation Using Bayesian Model Based Clustering'. The knowledge of his PhD project is successfully used in the day-to-day business of The SmartAgent Company.

²is Professor of Applied Bayesian Statistics at the Department of Methodology and Statistics in the Faculty of Social and Behavioural Sciences of Utrecht University. Besides Bayesian Statistics, his areas of expertise are, among others, model-based clustering for large data sets, regression, (repeated) measures, analysis of variance, multiple hypotheses testing and structural equation modelling.

Received 18 November 2008; Revised 18 November 2008.

Abstract

Data fusion, or combining multiple data sets into one data set, is not a new concept. Because of increasing desire for differentiated direct marketing strategies, however, it is getting more popular in marketing. This paper shows how marketing information can be fused to a company's customer database. Using a real marketing application, two traditional data fusion methods, polytomeous logistic regression and nearest neighbour algorithms, are compared with two model-based clustering approaches. Finally, the results are evaluated using internal and external criteria.